Epsilon substitution for transfinite induction (Q1778059)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Epsilon substitution for transfinite induction
scientific article

    Statements

    Epsilon substitution for transfinite induction (English)
    0 references
    0 references
    26 May 2005
    0 references
    The systems \(\text{PA}+\text{TI}(\prec)\) of first-order arithmetic extended by a scheme of transfinite induction on a recursive well-ordering \(\prec\) (of the order type \(\alpha\)) are often used in proof theory. \textit{T. Arai} [``Epsilon substitution method for \(\text{ID}_{1}(\Pi_{1}^{0}\vee{\Sigma} _{1}^{0})\)'', Ann. Pure Appl. Logic 121, 163--208 (2003; Zbl 1022.03040)] outlined a schema of an \(\varepsilon\)-substitution method for such systems and of a termination proof using ordinary assignment by a method of W. Ackermann. The author provides a cross-check via a termination proof using cut elimination following \textit{G. Mints} [``Gentzen-type systems and Hilbert's epsilon substitution method. I'', in: D. Prawitz et al. (eds.), Logic, methodology and philosophy of science IX. Proceedings of the ninth international congress of logic, methodology and philosophy of science, Uppsala, Sweden, August 7--14, 1991. Amsterdam: North-Holland. Stud. Logic Found. Math. 134, 91--122 (1994; Zbl 0835.03023)]. The main new feature of the formulation by T. Arai is a possibility to assign an arbitrary value \(n\) satisfying \(\varphi[n]\) to an \(\varepsilon\)-term \(\varepsilon x\varphi[x]\), not necessarily the \(\prec\)-least \(n\), since the latter is non-computable in general. A correction step corresponding to a critical formula \(\varphi[m]\to\varphi[\varepsilon x\varphi[x]]\) may change such a value \(n\) to \(m\prec n\). The main new feature compared to the termination proof of Mints [loc. cit.] is the treatment of cuts corresponding to the alternative: either \((\varepsilon x\varphi[x],?)\) (that is, \(\forall x\neg \varphi[x]\) and \(\varepsilon x\varphi[x]\) has a default value 0) or (non-default) \(\varepsilon x\varphi[x]=n\) for some \(n\). Redexes corresponding to this cut are situated over an \((\varepsilon x\varphi[x],?)\)-premise and called \(H\)-axioms. An \(H\)-axiom \(A\times H_{\varepsilon x\varphi[x],n}\) corresponds to \(\varepsilon x\varphi[x]=n\). In the reviewer's paper [loc. cit.] a reduction of such a redex used only the premise of the cut corresponding to \(\varepsilon x \varphi[x]=n\). In the present treatment cut elimination in premises for many \(m\preceq n\) is used in a way corresponding to the quantifier \((\forall m\prec x)\) in the progressivity condition \(\forall x(\forall y\prec x\,\, Px\to Py)\). Otherwise the schema of a proof is the same as in the reviewer's paper [loc. cit.]. For a given finite system \(E\) of critical \(\varepsilon\)-formulas, an \(\omega\)-derivation is constructed, cut-elimination theorem is proved, and it is noted that a cut-free derivation is finite and encodes a terminating \(\varepsilon\)-substitution process.
    0 references
    first-order arithmetic
    0 references
    transfinite induction
    0 references
    recursive well-ordering
    0 references
    epsilon-substitution method
    0 references
    termination proof
    0 references
    cut elimination
    0 references
    0 references

    Identifiers